How many images

[Pkay]

With in the PI documentation for 'Image Integration' there is a graph.
See attached screen shot.
It says 'beyond 30-50 images there is little to be gained.


I put this information to my local astro photography forum and there was strong disagreement.

There are many who take 100's of subs. and they claim there is a lot to be gained.

Who is right?

[RickS]

Peter,

The problem I have with that graph is that it makes it look like there's a single SNR curve for the whole image.  In reality, the curve looks different for different parts of the image and dim and bright areas behave quite differently.

There's a discussion of the 30 sub thing here: https://pixinsight.com/forum/index.php?topic=10461.0

I love Mike Schuster's comment: "Dim targets require more time/subs. Dim targets don't respect 30 sub limits."

Cheers,
Rick.

[sharkmelley]

Who is right?  Both are.

The graph correctly shows that as the number of images increases, each additional image has less and less effect.  On the other hand, for each doubling of the number of images, SNR increases by square root of 2.  There is no contradiction between the two.

I generally shoot hundreds of subs.  But they are short subs, 30seconds each.  400 subs is visibly better than 200 subs which is visibly better than 100 subs which is visibly better than 50 subs.

The only practical limit is how many hours of data you are prepared to devote to a final image.

Mark

[Pkay]

Thanks Rick, Mark

Have just finished reading the attached link.
Wow, what a can of worms!

At this point concluded the question is almost unanswerable. 

But I do like this quote regarding the 30 image limit: 'It over-simplifies things for the novice, confuses the experienced, and is ignored by the professional.’

[RickS]

At this point concluded the question is almost unanswerable. 

Went back and revisited my previous calculations and came to the same conclusion again   What matters is the total integration time, not the number of subs it is divided into. 30 sky limited subs of 10 minutes will give you the same SNR as 300 sky limited subs of 1 minute or 3000 sky limited subs of 6 seconds (you'll need a low noise camera for the 1 minute subs and a very low noise camera for the 6 second ones, of course.)  Should you stop at 30 x 6 second subs because of diminishing returns?  No, obviously not.  It's cheap to take a 6 second sub...

Cheers,
Rick.

[bulrichl]

Hi Rick,

your calculation is overoptimistic. It would be right, if NO residual FPN was left in the calibrated frames at all. In practice this is not achievable, even if you (very important) dither with sufficient large amount between frames and do the calibration right. I guess this was what Jon Rista wanted to express in his post ( https://pixinsight.com/forum/index.php?topic=10461.0 , response #4).

So the real-world result will be, that remaining FPN (after calibration) sets the definite limit, beyond which an improvement no longer is achievable. However, this limit is not fixed to a certain number of exposures. It depends critically on the quality of dithering and on the quality of the calibration.

Bernd

[RickS]

Bernd,

I'm aware that this is a gross oversimplification, but hopefully it makes the point anyway.

Cheers,
Rick.

[bulrichl]

Rick,

I agree that total exposure time is what matters if you follow best practice in image acquisition and processing. However, I wanted to stress that at some point residual FPN will set a definite limit to further improvement of SNR. This limit can be quite low, when e.g. one does not dither or takes an insufficient amount of calibration frames. In this case it would not pay to take a great many of light frames.

I agree as well that a certain number of exposures does not seem to be a useful criterion. It depends on the data and the processing.

Bernd

[RickS]

Good point, Bernd.  I hadn't considered that someone might be foolish enough to fail to follow best practice 

[bulrichl]

Not shure whether you got my point ("... residual FPN will set a definite limit to further improvement of SNR"). That holds generally.

Bernd

[RickS]

Not shure whether you got my point ("... residual FPN will set a definite limit to further improvement of SNR"). That holds generally.

Yes, I did, Bernd.  I also agree that the limit is much lower if you don't calibrate well, dither, etc.  However, the OP is not looking for a complex, nuanced answer.